Seung-Woo Baek

Drug resistance remains a significant barrier to achieving durable treatment responses. Traditionally, resistance has been attributed to genetic alterations and clonal selection. However, accumulating evidence suggests that early adaptation to therapy is often mediated by non-genetic state transitions. In this review, we propose a conceptual framework in which resistance emerges through therapy-driven molecular evolution in bladder cancer, characterized by three interconnected axes: non-genetic plasticity, metabolic reorganization, and tumor microenvironment remodeling. Using the Gemcitabine-Resistant Cell (GRC) model as a temporal reference system, we describe a stepwise transition from drug-sensitive states dominated by proliferation to survival-optimized resistant states through a growth–survival trade-off. Early adaptive phases are marked by the attenuation of cell-cycle and glycolytic programs, increased epigenetic flexibility, and metabolic rewiring involving mitochondrial and lipid-associated pathways. Later phases involve the reinforcement of resistance through extracellular matrix remodeling, developmental and stress-response signaling, and immunometabolic interactions within the tumor microenvironment, including adenosine- and lipid-associated mediators. Projecting the GRC score onto a clinical bladder cancer cohort further suggests that these evolutionary patterns may also be reflected in patient tumors. Overall, this framework supports a temporally structured view of chemoresistance and highlights opportunities to therapeutically target transitional adaptive states before resistance becomes stabilized.

Background Muscle-invasive bladder cancer (MIBC) is a clinically aggressive and heterogeneous disease with variable treatment responses. Transcriptome-based classifications, such as the Chemoresistance-Motility (CrM) signature, are valuable for understanding therapeutic resistance, but their clinical use is often hindered by high cost and tissue requirements. This study explores an alternative, scalable approach using deep learning analysis of whole slide images (WSIs). Objective We aimed to evaluate whether patch-level predictions from deep learning models applied to WSIs can accurately predict transcriptome-derived CrM subtypes and reflect tumor microenvironment (TME) characteristics in MIBC. Methods We analyzed 192 WSIs from 152 TCGA-BLCA patients. A pretrained deep learning model (densenet169-kather100k) was used to classify eight distinct tissue types per patch. A key histological metric, the SAM-to-DNT ratio, which represents the ratio of stromal, adipose, and smooth muscle (SAM) tissue types to debris, normal, and tumor epithelium (DNT) tissue types, was derived from these proportions. A random forest model was then trained on these features to predict CrM subtypes. Results The WSI-derived SAM-to-DNT ratio showed a strong positive correlation with the CrM score (R = 0.453) and transcriptome-based TME scores, such as cancer-associated fibroblasts (R = 0.398). Our random forest model successfully classified CrM subtypes with a balanced accuracy of 0.75, outperforming other algorithms. Feature importance analysis identified adipose tissue (ADI) and tumor epithelium (TUM) as the most predictive features for CrM status. Conclusions Deep learning analysis of routine histological WSIs can serve as a practical, low-cost surrogate for molecular profiling, effectively capturing transcriptomic subtypes associated with chemoresistance in MIBC. This approach provides a viable method for patient stratification and establishes a foundation for future multi-modal precision oncology applications.

Approximately 75% of bladder cancer cases originate as non-muscle-invasive bladder cancer (NMIBC). Despite initial diagnosis, NMIBC commonly recurs, with up to 45% advancing to muscle-invasive bladder cancer (MIBC) and metastatic disease. Treatment for high-risk NMIBC typically includes procedures like transurethral resection and, depending on recurrence risk, intravesical chemotherapy or immunotherapy such as Bacillus Calmette–Guérin (BCG). However, persistent shortages of BCG necessitate alternative first-line treatments. We aim to use a multi-gene signature in high-risk NMIBC patients to determine whether patients may benefit from immune checkpoint inhibitors (ICIs) as an alternative to BCG and to evaluate their clinical utility. The multi-gene signature obtained from the three independent NMIBC cohorts was applied to stratify the UROMOL2016 cohort (n = 476) using consensus clustering. Each subtype was distinguished by biological pathway analysis. Validation analysis using a machine learning algorithm was performed in six independent cohorts including the BRS (n = 283) cohort treated with BCG and the IMvigor210 (n = 298) clinical trials treated with PD-L1 inhibitors. Based on consensus cluster analysis, NMIBC patients in the UROMOL2016 cohort were classified into three classes exhibiting distinguished characteristics, including DNA damage repair (DDR). Survival analysis showed that the NMIBC-DDR class had the highest rates of disease progression (progression-free survival, p = 0.002 by log-rank test) in the UROMOL cohort and benefited from BCG and ICIs (respectively, p = 0.02 and p = 0.03 by log-rank test). This study suggests that the multi-gene signature may have a role in identifying high-risk NMIBC patients and improving the responsiveness of ICIs. Additionally, we propose immunotherapy as a new first-line treatment for patients with high-risk NMIBC because of the shortage of BCG supply. It is important to help more patients prioritize cancer immunotherapy.

Chemotherapy resistance is an obstacle to cancer therapy and is considered a major cause of recurrence. Thus, understanding the mechanisms of chemoresistance is critical to improving the prognosis of patients. Here, we have established a stepwise gemcitabine-resistant T24 bladder cancer cell line to understand the molecular mechanisms of chemoresistance within cancer cells. The characteristics of the stepwise chemoresistance cell line were divided into 4 phases (parental, early, intermediate, and late phases). These four phase cells showed increasingly aggressive phenotypes in vitro and in vivo experiments with increasing phases and revealed the molecular properties of the biological process from parent cells to phased gemcitabine-resistant cell line (GRC). Taken together, through the analysis of gene expression profile data, we have characterized gene set of each phase indicating the response to anticancer drug treatment. Specifically, we identified a multigene signature (23 genes including GATA3, APOBEC3G, NT5E, MYC, STC1, FOXD1, SMAD9) and developed a chemoresistance score consisting of that could predict eventual responsiveness to gemcitabine treatment. Our data will contribute to predicting chemoresistance and improving the prognosis of bladder cancer patients.

Contrary to many reports that antiplatelet agents inhibit cancer growth and metastasis, new solid tumors have been reported in patients receiving long-term antiplatelet therapy. We investigated the effects of these agents directly on cancer cells in the absence of platelets to mimic the effects of long-term therapy. When four antiplatelet agents (aspirin, clopidogrel, prasugrel, and ticagrelor) were administered to colon cancer cells, cancer cell proliferation was inhibited similarly to a previous study. However, surprisingly, when cells were treated with a purinergic P2Y12 inhibitor (purinergic antiplatelet agent), the motility of the cancer cells was significantly increased. Therefore, gene expression profiles were identified to investigate the effect of P2Y12 inhibitors on cell mobility, and Serpin family 1 (SERPINE1) was identified as a common gene associated with cell migration and cell death in three groups. Antiplatelet treatment increased the level of SERPINE1 in cancer cells and also promoted the secretion of SERPINE1 into the medium. Increased SERPINE1 was found to induce MMP1 and, thus, increase cell motility. In addition, an increase in SERPINE1 was confirmed using the serum of patients who received these antiplatelet drugs. With these results, we propose that SERPINE1 could be used as a new target gene to prevent the onset and metastasis of cancer in patients with long-term antiplatelet therapy.

DNA repair defects are important factors in cancer development. High DNA repair activity can affect cancer progression and chemoresistance. DNA double-strand breaks in cancer cells caused by anticancer agents can be restored by non-homologous end joining (NHEJ) and homologous recombination repair (HRR). Our previous study has identified E2F1 as a key gene in bladder cancer progression. In this study, DNA repair genes related to E2F1 were analyzed, and RAD54L involved in HRR was identified. In gene expression analysis of bladder cancer patients, the survival of patients with high RAD54L expression was shorter with cancer progression than in patients with low RAD54L expression. This study also revealed that E2F1 directly binds to the promoter region of RAD54L and regulates the transcription of RAD54L related to the HRR pathway. This study also confirmed that DNA breaks are repaired by RAD54L induced by E2F1 in bladder cancer cells treated with MMC. In summary, RAD54L was identified as a new target directly regulated by E2F1. Our results suggest that, E2F1 and RAD54L could be used as diagnostic markers for bladder cancer progression and represent potential therapeutic targets.

Recent investigations reported that some subtypes from the Lund or The Cancer Genome Atlas (TCGA) classifications were most responsive to PD-L1 inhibitor treatment. However, the association between previously reported subtypes and immune checkpoint inhibitor (ICI) therapy responsiveness has been insufficiently explored. Despite these contributions, the ability to predict the clinical applicability of immune checkpoint inhibitor therapy in patients remains a major challenge. Here, we aimed to re-classify distinct subtypes focusing on ICI responsiveness using gene expression profiling in the IMvigor 210 cohort (n = 298). Based on the hierarchical clustering analysis, we divided advanced urothelial cancer patients into three subgroups. To confirm a prognostic impact, we performed survival analysis and estimated the prognostic value in the IMvigor 210 and TCGA cohort. The activation of CD8⁺ T effector cells was common for patients of classes 2 and 3 in the TCGA and IMvigor 210 cohort. Survival analysis showed that patients of class 3 in the TCGA cohort had a poor prognosis, while patients of class 3 showed considerably prolonged survival in the IMvigor 210 cohort. One of the distinct characteristics of patients in class 3 is the inactivation of the TGFβ and YAP/TAZ pathways and activation of the cell cycle and DNA replication and DNA damage (DDR). Based on our identified transcriptional patterns and the clinical outcomes of advanced urothelial cancer patients, we constructed a schematic summary. When comparing clinical and transcriptome data, patients with downregulation of the TGFβ and YAP/TAZ pathways and upregulation of the cell cycle and DDR may be more responsive to ICI therapy.